To identify significant predictors of device-related infections, we performed a prospective, nonrandomized analysis of our experience with vascular access devices over a 2-year period in a pediatric oncology population. Variables analyzed included: (1) age at placement, (2) sex, (3) underlying disease, (4) type of device used (catheter v port), and (5) total white blood cell count at placement. Quantitative microbiologic criteria were used for diagnosis of bacteremia while clinical and microbiologic criteria were used in diagnosis of tunnel/port/site infections. During the study period a total of 351 devices, comprising 78,159 days in situ, were placed and data for univariate and multivariate analysis were available on 271 (77%). The mean age at placement was 7.2 +/- 4.7 years for catheters and 9.5 +/- 4.8 years for implantable devices (P less than or equal to .01). Significant predictors of device-related infections in univariate analysis were type of device (P less than or equal to .0001) and age (P less than or equal to .002). External catheters and age less than or equal to 7 years were associated with increased risk of infection. Underlying disease had a marginal effect on the infection rate (P = .08). In multivariate analysis, device type (P less than or equal to .0001) and age (P less than or equal to .002) continued to affect infections, whereas underlying disease demonstrated only a borderline effect (P = .14). We conclude that device type and age significantly affect the rate of device-related infections. These data support increased use of implantable devices in pediatric oncology patients.